Chemical Synthesis of Affibody Molecules for Protein Detection and Molecular Imaging

(1)

Protein Detection and Molecular Imaging

TORUN EKBLAD

(2)

Stockholm 2008

Royal Institute of Technology School of Biotechnology AlbaNova University Center SE-106 91 Stockholm Sweden

Printed by Universitetsservice US-AB Drottning Kristinas väg 53B

SE-100 44 Stockholm Sweden

ISBN 978-91-7415-152-7 TRITA BIO-Report 2008:22 ISSN 1654-2312

(3)

Torun Ekblad (2008): Chemical Synthesis of Affibody Molecules for Protein Detection and Molecular Imaging. School of Biotechnology, Royal Institute of Technology (KTH), Stockholm, Sweden.

Abstract

Proteins are essential components in most processes in living organisms. The detection and quantification of specific proteins can be used e.g. as measures of certain physiological conditions, and are therefore of great importance. This thesis focuses on development of affinity-based bioassays for specific protein detection. The use of Affibody molecules for specific molecular recognition has been central in all studies in this thesis. Affibody molecules are affinity proteins developed by combinatorial protein engineering of the 58-residue protein A-derived Z domain scaffold. In the first paper, solid phase peptide synthesis is investigated as a method to generate functional Affibody molecules. Based on the results from this paper, chemical synthesis has been used throughout the following papers to produce Affibody molecules tailored with functional groups for protein detection applications in vitro and in vivo.

In paper I, an orthogonal protection scheme was developed to enable site-specific chemical introduction of three different functional probes into synthetic Affibody molecules. Two of the probes were fluorophores that were used in a FRET-based binding assay to detect unlabeled target proteins. The third probe was biotin, which was used as an affinity handle for immobilization onto a solid support. In paper II, a panel of Affibody molecules carrying different affinity handles were synthesized and evaluated as capture ligands on microarrays.

Paper III describes the synthesis of an Affibody molecule that binds to the human epidermal growth factor receptor type 2, (HER2), and the site-specific incorporation of a mercaptoacetyl- glycylglycylglycine (MAG3) chelating site in the peptide sequence to allow for radiolabeling with^99mTc. The derivatized Affibody molecule was found to retain its binding capacity, and the

99mTc-labeling was efficient and resulted in a stable chelate formation. ^99mTc-labeled Affibody molecules were evaluated as in vivo HER2-targeting imaging agents in mice. In the following studies, reported in papers IV-VI, the ^99mTc-chelating sequence was engineered in order to optimize the pharmacokinetic properties of the radiolabeled Affibody molecules and allow for high-contrast imaging of HER2-expressing tumors and metastatic lesions. The main conclusion from these investigations is that the biodistribution of Affibody molecules can be dramatically modified by amino acid substitutions directed to residues in the MAG3-chelator.

Finally, paper VII is a report on the chemical synthesis and chemoselective ligation to generate a cross-linked HER2-binding Affibody molecule with improved thermal stability and tumor targeting capacity.

(4)

List of publications

This thesis is based on the following papers, which are referred to in the text by their Roman numerals (I-VII). The seven papers are found in the appendix.

I Engfeldt T., Renberg B., Brumer H., Nygren P-Å. and Eriksson Karlström A.

(2005) Chemical synthesis of triple-labelled three-helix bundle binding proteins for specific fluorescent detection of unlabelled protein. ChemBioChem 6: 1043-50

II Renberg B., Shiroyama I., Engfeldt T., Nygren P-Å. and Eriksson Karlström A.

(2005) Affibody protein capture microarrays: synthesis and evaluation of random and directed immobilization of affibody molecules. Anal Biochem 341:334-43

III Engfeldt T., Orlova A., Tran T., Bruskin A., Widström C., Eriksson Karlström A.

and Tolmachev, V. (2007) Imaging of HER2-expressing tumours using a synthetic Affibody molecule containing the ^99mTc-chelating mercaptoacetyl-glycyl-glycyl-glycyl (MAG3) sequence. Eur J Nucl Med Mol Imaging 34:722-33

IV Engfeldt T., Tran T., Orlova A., Widström C., Feldwisch J., Abrahmsén L., Wennborg A., Eriksson Karlström A. and Tolmachev V. (2007) ^99mTc-chelator engineering to improve tumour targeting properties of a HER2-specific Affibody molecule. Eur J Nucl Med Mol Imaging 34:1843-53

V Tran T., Engfeldt T., Orlova A., Sandström M., Feldwisch J., Abrahmsén L., Wennborg A., Tolmachev V. and Eriksson Karlström A. (2007). ^99mTc-maEEE- ZHER2:342, an Affibody molecule-based tracer for the detection of HER2 expression in malignant tumors. Bioconjug Chem 18:1956-64

VI Ekblad T., Tran T., Orlova A., Widström C., Feldwisch J., Abrahmsén L., Wennborg A., Eriksson Karlström A. and Tolmachev V. (2008) Development and preclinical characterisation of ^99mTc-labelled Affibody molecules with reduced renal uptake. Eur J Nucl Med Mol Imaging in press DOI 10.1007/s00259-008-0845-7

(5)

VII Ekblad T., Tolmachev V., Orlova A., Lendel C., Abrahmsén L. and Eriksson Karlström A. (2008) Synthesis and chemoselective intramolecular cross-linking of a HER2-binding Affibody. Submitted

Related publications not included in this thesis

Tran T., Engfeldt T., Orlova A., Widström C., Bruskin A., Tolmachev V. and Eriksson Karlström A. (2007) In vivo evaluation of cysteine-based chelators for attachment of ^99mTc to tumor-targeting Affibody molecules. Bioconjug Chem 18:549-58

Orlova A., Tran T., Widström C., Engfeldt T., Eriksson Karlström A. and Tolmachev V.

(2007) Pre-clinical evaluation of [¹¹¹In]-benzyl-DOTA-ZHER2:342, a potential agent for imaging of HER2 expression in malignant tumors. Int J Mol Med 20:397-404

Elfström N., Juhasz R., Sychugov I., Engfeldt, T., Eriksson Karlström, A. and Linnros J.

(2007) Surface charge sensitivity of silicon nanowires: size dependence. Nano Lett 7:2608-2612

Tran A. T., Ekblad T., Orlova A., Sandström M., Feldwisch J., Wennborg A., Abrahmsén L., Tolmachev V. and Eriksson Karlström A. (2008) Effects of lysine-containing mercaptoacetyl- based chelators on the biodistribution of ^99mTc-labeled anti-HER2 Affibody molecules.

Bioconjug Chem in press

(6)

INTRODUCTION

1. Proteins

Proteins are polymeric biomolecules, assembled from a repertoire of 20 amino acids. An amino acid consists of an amino group, a carboxyl group and a distinctive side chain, all of which are bonded to a central ơ-carbon atom. Condensation reactions between the amino and carboxyl groups result in the formation of a peptide bond (Figure 1). Polymeric sequences of amino acids joined by peptide bonds are called peptides, or polypeptides. The peptide backbone is not very reactive chemically, but presents potent donors and acceptors for hydrogen bond formations. Hydrogen bonding between CO- and NH-groups in the peptide backbone makes polypeptides fold into secondary structures, such as alpha helices or beta sheets (Pauling, 1951a, Pauling, 1951b). Additional hydrogen bonds, as well as weak, reversible interactions, such as electrostatic and van der Waals interactions, with the amino acid side chains assist in the folding of proteins into three-dimensional structures. The spatial

(10)

The amino acid building blocks are characterized by their different side chains, which all differ in size, shape, charge, hydrogen-bonding capacity and chemical reactivity. The order in which the amino acids are assembled is called the primary sequence and is of great importance since it dictates the structure and thereby also the function of the folded protein (Anfinsen, 1973).

With few exceptions, all proteins in all species are constructed from the same set of 20 amino acids, and differ only in the number of amino acids, and the sequence in which the various amino acids occur. The average eukaryotic protein is linear and unbranched and consists of 280 amino acids (Creighton, 1997). With 20 amino acids to choose from, and the unrestricted freedom to organize them sequentially, a huge number of primary sequences can be generated.

The diversity and versatility of the 20 amino acids allows for an overwhelming range of biological functions to be generated.

Figure 1. Formation of a peptide bond between two amino acids

Indeed, proteins have many different biological functions. Most chemical reactions in biological systems are catalyzed by proteins with enzymatic functions. Proteins play essential roles in the regulation of cell growth and differentiation. In particular protein hormones are important regulators. Proteins also serve important functions in transport and storage of small molecules and ions, such as myoglobin transportation of oxygen and ferritin for iron storage.

Many proteins, such as keratin or collagen, provide mechanical support. Antibodies is another important type of proteins that serve to protect its host from foreign substances.

In addition to the functionalities that are provided in the amino acid side chains, many proteins are subjected to various chemical modifications during or after assembly of the peptide chain.

Proteolytic processing, alterations of the chain termini and coupling of functional groups to amino acid side chains are examples of modifications that are naturally introduced to tailor proteins for improved performance in certain functions. Proteins can be glycosylated to alter the specificity in their interactions with other biomolecules, to improve the solubility, or to lengthen the biological half-life. Lipids can be attached in order to tether proteins to membranes. Phosphorylation of serine, threonine, and tyrosine residues is an important

(11)

regulator of protein activity, and hydroxylation is an essential modification for folding and assembly of collagen.

Important classes of proteins are those that have the ability to interact specifically with other proteins. This thesis focuses on such affinity proteins, and methods to utilize molecular recognition for specific protein detection. In order to develop highly sensitive bioassays, the affinity proteins often need to be modified with functional groups that improve their performance. Chemical synthesis is a production route available for small proteins, and offers a means to introduce such functional probes site-specifically to modify the function of the protein.

(12)

2. Affinity proteins

A specific interaction between two molecules is termed molecular recognition. In nature, proteins and peptides are involved in various non-covalent interactions with other biomolecules, such as antigen-antibody, receptor-ligand, DNA-protein, sugar-lectin and RNA- ribosome interactions. Such natural examples of molecular recognition have been transferred to applications in biotechnology and medicine, including for example bioseparation using lectins as ligands for carbohydrate-recognition, and development of diagnostic assays such as skin tests to detect bacterial infections (Campos-Neto, 2001) or allergies. Natural protein interactions have also been used in clinical therapy for regulation of signaling pathways, as in the use of insulin receptor interactions to regulate glucose homeostasis, or for targeted delivery of therapeutic payloads, such as the use of growth hormones for delivery of toxins or radionuclides to epidermal growth factor receptors (Sundberg, 2003). In addition to protein interaction systems derived from natural interaction pairs, development of engineering strategies to generate novel affinity proteins for defined biomolecular recognition has been an intensive research field since the 1990s.

Depending on the intended application, different requirements are set for the affinity proteins.

Most importantly, an affinity reagent has to bind with adequate affinity and selectivity to its target or group of targets. For applications where the bound protein should be recovered, it is essential to have a possibility to break the interaction by addition of suitable agents. The route by which the affinity protein can be produced is an important consideration. Ideally, it should be possible to obtain large quantities in a cost-efficient manner. For many applications it is also necessary to have a production route that allows for the introduction of chemical modifications. Stability to chemical, physical and enzymatical challenges are important issues, as are biophysical parameters such as size and polarity. For in vivo use, the species from which the affinity proteins originate has to be regarded for immunogenicity reasons.

The most widely used affinity proteins for biotechnological and medical applications are antibodies. Antibodies have played an essential role in the development of most applications that rely on molecular recognition. The importance of this class of affinity proteins can be illustrated by the number of Nobel Prizes earned in the field of antibody research, which is no less than seven (http://www.nobelprize.org). There are currently 32 antibodies and antibody- derived molecules approved by the US FDA to be used in the clinic for diagnostics and

(13)

therapy (Leader, 2008). However, certain shortcomings of antibodies have been identified, which have generated an interest in development of alternative reagents for molecular recognition. Using the conceptual architecture of antibodies and the principles by which variability can be created, strategies have evolved to develop new classes of affinity proteins, including fragments derived from antibodies and affinity proteins developed from alternative scaffolds. Such molecules have come forward as complement or alternatives to antibodies in various settings where characteristics different from those of antibodies are useful.

2.1 Antibodies

The existence and survival of an organism rely heavily on its capacity to protect itself from foreign invaders. The immune system of humans, and higher vertebrates, offers an ingenious source of diversification and means for selection of high affinity antibodies that recognize most imaginable alien substances. Antibodies have evolved to serve two functions, to specifically interact with foreign substances in the body, and to recruit effector functions of the immune system.

A human antibody consists of four polypeptide chains, two identical light chains (L) and two identical heavy chains (H), each of which is organized into multiple globular domains of hydrogen-bonded, antiparallel Ƣ-strands with connecting loops and stabilizing disulfide bonds.

The domains all have a similar fold, called the immunoglobulin fold, but differ in terms of variability, and are divided into constant domains with strictly conserved sequences and variable domains with high sequence diversity. The antibody is Y-shaped, with constant domains of the heavy chains forming the stem, and each arm formed by two constant domains (CH1 and CL) and two variable domains (VH and VL) at the very tip (Figure 2).

The sequence of the constant regions of the heavy chain can only vary according to certain

(14)

Figure 2. Schematic representation of an IgG antibody

In the structures of immunoglobulins, high variability is observed in the loops that connect the Ƣ-strands, both in terms of length and amino acid sequence. Three loops each in the variable domains of the heavy and light chain respectively are hypervariable and called complementarity determining regions, (CDRs). The rest of the variable domains are known as framework regions and act as a scaffold that brings the six CDRs together to form the surface that is responsible for antigen binding. The great diversity in terms of composition and topology (flat, crevice-like or protruding) that the binding surface of antibodies can have allows for highly specific interactions with ligands of varying size and shape, from small haptens to large macromolecules. By different mechanisms occurring within the immunoglobulin-encoding genes in maturing B-cells (i.e. random gene segment shuffling, frame-shifting, nucleotide insertions, and somatic mutations), a vast diversity of binding specificity can be generated from a limited number of original genes. Individual members of this library of antibody variants is displayed by B-cells in the circulation, which upon encountering its cognate antigen, and after receiving stimulatory signals from helper T-cells, will start to differentiate into plasma cells, secreting large amounts of a specific antibody. The fine specificity and affinity of the antibody

(15)

can be further improved during the life of the B-cell lineage via somatic hypermutations in the CDR regions (Goldsby, 2003).

It has been suggested that basically any macromolecular structure can be immunogenic, if presented properly to the immune system (Benjamin, 1984). This fact has been employed in efforts to generate antibodies with desired specificities. By immunizing laboratory animals with an antigen of interest, an immune response can be raised, generating a pool of antibodies recognizing different sites (epitopes) of the antigen. This heterogeneous ensemble of molecules is termed polyclonal antibodies and has proved to be useful as affinity reagents in many biotechnological applications in vitro such as ELISA (Engvall, 1971), microarrays (Bertone, 2005), and affinity proteomics (Uhlén, 2005). However, immunizations may not generate an entirely reproducible response, which might be an issue in regulated manufacturing procedures. Also, for in vivo applications it is desirable to use reagents of more well-defined character.

The reproducible production of defined antibodies became possible with the development of the hybridoma-technology (Kohler, 1975). By fusion of mouse myeloma and mouse spleen cells (i.e. antibody-producing B-cells) from an immunized donor, individual in vitro-dividing cell-lines were generated that each produced a particular antibody variant, a monoclonal antibody. The development of the hybridoma technology came to revolutionize the use of antibodies and was later awarded the Nobel Prize. Some limitations of monoclonal antibody production by the hybridoma technology are that it is laborious, time-consuming and cannot be realized in a high-throughput fashion.

Important early steps towards a by-pass of both immunization and the hybridoma technology for obtaining antibodies were provided when recombinant production of antibodies was reported for bacterial cells. The first recombinant production of complete H- and L-chains in Escherichia coli was described in 1984 (Cabilly, 1984). However, the proteins were obtained as insoluble aggregates and refolding efforts resulted in poor recovery of activity. In 1988, two

(16)

The successful cloning of antibody genes also allowed for genetic engineering, which could be used to address some of the issues regarding immunogenicity caused by the murine origin of monoclonal antibodies obtained by the hybridoma technology. By using genetic engineering, chimeric antibodies with the constant region of human origin and the variable domains carrying the antigenic specificity derived from mouse have been generated (Morrison, 1984).

Another important progress was the approach to graft only the CDRs from the mouse clone onto a human antibody framework to generate humanized antibodies (Jones, 1986). Thereby, all constant domains are encoded by human genes, resulting in reduced immunogenicity and conservation of the biological effector functions, thus rendering the antibodies likely to trigger immune system reactions such as the complement activation or Fc receptor binding.

Cloning of antibody genes and the realization of recombinant production routes have been crucial to allow for different types of protein engineering of antibodies, such as site-directed mutagenesis or generation of fusion proteins. Furthermore, it inspired the development of genetically engineered antibody-derived fragments, which have been found to be better suited both for expression and in vitro selection. Small recombinant antibody fragments and engineered variants thereof have emerged as alternatives to full-sized antibodies in several applications where size and complexity become decisive issues.

2.2 Antibody fragments

For many applications, high-yield production and solubility of the affinity protein are critical factors. Furthermore, small size may prove essential to achieve a desired biodistribution in vivo, for applications such as molecular imaging. Size reduction can also be used as a strategy to reduce the immunogenicity of an affinity protein of non-human origin. Many attempts to reduce the size of antibodies while retaining its antigen-binding properties have been reported, some of which are described below.

The initial steps towards dissecting the antibody molecule were made by enzymatic digestion, revealing fragments with retained antigen-binding, namely Fab and F(ab)’2 (Porter, 1959). The Fab fragment has a size of 54 kDa and is made up of the complete light chain and the variable domain and the first constant domain from the heavy chain, which form the arm of the Y- shaped antibody structure (Figure 2). The difference between the fragments Fab and Fab’ is that the latter also includes the free cysteine residue that originally is included in the first hinge disulfide with the other heavy chain. F(ab)’2 is the disulfide-bonded dimer of two Fab’

(17)

fragments. Fab fragments were the first functional antibody-derived fragments to be successfully produced in bacterial expression systems (Better, 1988), (Skerra, 1988).

The smaller Fv fragment consists of only the variable domains of the heavy and light chain (VH

and VL). The domains associate in a non-covalent manner, which have been difficult to accomplish in recombinant expression settings. This has been by-passed with the development of single chain Fv fragments, scFv (27 kDa) (Bird, 1988), where a flexible peptide loop that links the variable domains allows for recombinant production of a single polypeptide chain and also brings a higher stability to the molecule. The scFv’s have found many applications as affinity reagents, such as in microarrays (Wingren, 2005). Dimers of scFv’s, called diabodies, have also been generated (Holliger, 1993). By using a linker that is too short to allow for interactions between domains on the same chain, dimeric molecules are forced to assemble by pairing with the complementary domain of another chain. The fact that diabodies have two functional antigen binding sites brings increased apparent affinity (avidity) compared to the monomer. Alternatively, this feature can be used to create molecules with dual binding affinities.

The variable domains VH and VL might be used individually as small (15 kDa) antigen binding units (Ward, 1989). However, a large hydrophobic surface, which originally makes up the interaction surface of the two domains, is exposed upon their separation, resulting in poor solubility of the individual fragments. Efforts to minimize the exposed hydrophobic interface by site-directed mutagenesis have resulted in successful production of such binders. In nature, a unique class of antibodies devoid of the light chain and with antigen-binding sites composed of a single variable domain has been identified in camelids (Hamers-Casterman, 1993) and cartilaginous fish (Greenberg, 1995). These represent the smallest available antigen-binding fragments (15 kDa) derived from antibodies.

The successful cloning and expression of antibody genes in bacterial cells and the progress made in the design of antibody fragments, was followed by the development of methods to

(18)

into the phage genome (Smith, 1985). Genes for complete antibody V-domains have been amplified by PCR and the corresponding proteins successfully displayed on phages so that clones with binding activities to a desired antigen could be selected using affinity chromatography (McCafferty, 1990). Cloning the repertoire of antigen-binding fragments from an immunized donor into a phage display vector and selection of antigen-specific clones has since become a routine method (Winter, 1994) (Hoogenboom, 1998). Other in vitro selection systems such as mRNA display, ribosome display, and cell surface display on different types of organisms, such as yeast, and gram-positive bacteria, are being used for the same purpose and are reviewed by Lin and Cornish (Lin, 2002).

2.3 Alternative scaffolds

One inconvenient feature related to antibodies is that they are difficult to express recombinantly as correctly folded and biologically active proteins at high yields in bacterial expression systems. With powerful in vitro selection systems at hand, the use of alternative scaffolds, characterized by a more easily folded framework than the immunoglobulin, could be considered for harboring novel binding sites. The use of such alternative scaffolds can bring advantages such as favorable expression routes in terms of yield, stability, and folding as well as attractive biophysical characteristics and might be a way to realize some applications where antibodies are not an option. Examples of such applications are interactions with immuno- evasive clefts and grooves in a target protein (Stijlemans, 2004), and co-crystallization (Högbom, 2003), (Binz, 2004).

Rational de novo design of peptides or proteins with specific binding capacity is a very difficult task. However, a completely different route to developing proteins with new functions was taken in 1990 when it was reported that novel specific ligands could be selected from combinatorial libraries of short peptides with randomly generated sequences without any prior knowledge about the nature of the interaction surface (Cwirla, 1990), (Devlin, 1990), (Scott, 1990). In efforts to overcome the poor binding affinities of the flexible peptides, conformationally constrained peptides were applied (O'Neil, 1992), (Koivunen, 1995), and the idea to use already folded proteins for development of novel affinity reagents evolved.

Scaffold engineering involves the introduction of a novel binding specificity to a structural framework made up of a folded protein. This can be accomplished either by grafting a peptide sequence with a known function onto the scaffold, or by site-specific mutations of surface-

(19)

exposed residues in the scaffold to replace a parental binding specificity or to create a novel binding site. More than 50 different scaffolds have been suggested as alternative affinity proteins, and the topic has been thoroughly reviewed by others (Nygren, 2004), (Binz, 2005), (Hey, 2005), (Hosse, 2006), (Skerra, 2007) The following sections serve to present some representative examples and highlight the Affibody scaffold, which has been the focus of the studies that are presented in this thesis. Affinity proteins based on alternative scaffolds differ widely in their structural architecture and the way in which they interact with their target molecules. A classification based on the properties of the interaction surface results in the following classes; single loops on a rigid framework (chapter 2.3.1), several loops forming a contiguous surface (chapter 2.3.2) and engineered interfaces resting on a secondary structure (chapter 2.3.3). Some examples of affinity proteins derived from alternative scaffolds are summarized in Table 1.

2.3.1 Single loops on a rigid framework

In their simplest form, engineered affinity proteins consist of a single loop peptide with randomized sequence grafted onto a stable protein framework. Kunitz domains are a class of serine protease inhibitors of such design. Some 60 amino acids make up a stable fold, presenting the protease binding site as an exposed loop with varying sequence. By randomizing residues in the loop, libraries have been generated from which protease inhibitors with novel specificities have been selected using phage display (Dennis, 1994, Nygren, 2004). The main application of Kunitz domains has so far been as protease inhibitors, such as the DX-88 inhibitor of plasma kallikrein, which is now in clinical trials for treatment of hereditary angioedema (Williams, 2003).

Cystine-knot miniproteins, or knottins, represent another family of small (25-35 aa) proteins that has been used as alternative scaffolds for peptide loop grafting in combinatorial

(20)

was shown to be amenable to cell surface display on E.coli. Another member of the knottin family is the cellulose–binding domain, CBD, from fungal cellobiohydrolase, which has been subjected to combinatorial engineering to generate binders to alpha amylase (Lehtiö, 2000) and alkaline phosphatase (Smith, 1998). Furhtermore, knottins have been engineered to generate chimeric molecules that can accommodate multiple binding specificities on different faces of the scaffold (Le Nguyen, 1989). Another interesting feature of these small affinity proteins is that they can be produced by chemical synthesis, in addition to recombinant production. The head-to-tail cyclic cystine-knot protein Kalata BI has been synthesized and characterized in terms of folding pathways. It was discovered that fully deprotected peptides were able to fold independently with correct organization of the cystine-knot motif. However, cyclization of the peptide chain prior to oxidation of the thiol groups was found to increase the propensity for correct folding (Daly, 1999). As an alternative, a semisynthetic strategy has been applied for production of cyclic cystine-knot proteins, where recombinant production of a linear precursor peptide allowed for efficient folding and oxidation, and the head-to-tail cyclization was performed by chemical synthesis (Avrutina, 2008).

Another example of scaffolds that gain novel properties by variations of amino acids in a single continuous loop is thioredoxin. This small enzyme (108 aa) displays a disulfide- constrained solvent-exposed peptide loop in its active site that has been replaced by a 20 residue randomized sequence to generate molecules with novel binding capacities. Examples of such thioredoxin-derived peptide aptamers are those designed to interfere with the intracellular protein interactions of cyclin-dependent kinase (Cdk2) (Cohen, 1998) and the tyrosine kinase domain of ErbB2 (Kunz, 2006).

2.3.2 Several loops forming a contiguous surface

The description of this category of affinity proteins also fits the architecture of antibodies, in which pairs of variable Ig domains bring the six CDR loops together to form the antigen binding surface. The fibronectin type III domain (FN3) has a structure similar to that of an antibody VH domain but with only seven Ƣ-strands and no disulfide bonds. The FN3 framework provides several suitable properties such as being small (94 aa), monomeric, and with a ligand binding site formed by surface loops, and has been used as a scaffold to develop novel binding proteins (Koide, 1998) called monobodies or AdNectins. Targets to which FN3- variants have been selected include vascular endothelial growth factor receptor 2 (VEGFR2)

(21)

(Getmanova, 2006), TNF-ơ (Xu, 2002), the SH3 domain of human c-Src (Karatan, 2004) and the estrogen receptor (Koide, 2002).

The lipocalins constitute another class of affinity proteins with a structurally conserved Ƣ–

sandwich framework, which display loop regions with high variability in both length and sequence. The folding motif of the lipocalins is a Ƣ-barrel, made up of eight antiparallel Ƣ- sheets, with one end closed by densely packed hydrophobic residues and the other end forming an open ligand-binding pocket, flanked by four extended loops. Lipocalins with novel binding specificities, called anticalins (20 kDa), have been generated by random mutagenesis directed to sites in the loop regions. The anticalin scaffold allows for high affinity interactions with small hapten molecules as well as large proteins. The bilin-bindig protein BBP has been engineered to selectively interact with fluorescein (Beste, 1999) and digoxigenin (Schlehuber, 2000), whereas lipocalins of human origin have been utilized for development of anticalins for intended medical applications, such as an apolipoprotein D-derived hemoglobin-binder (Vogt, 2004) and a human neutrophil lipocalin engineered for specific interaction with CTLA-4 (Schlehuber, 2005).

2.3.3 Engineered interfaces resting on a secondary structure

This class of affinity proteins is derived from scaffolds that provide a secondary structure element with exposed side chains that have potential to be involved in molecular recognition.

One such scaffold is represented by repeat proteins. In essence, repeat proteins are built up of structural units that are stacked into an elongated domain with a continuous binding surface.

Repeat proteins are abundant in many species and are known to mediate protein-protein interactions (Kobe, 2000). Scaffolds have been designed based on consensus sequences of the repeated motifs in e.g. ankyrin repeats (Forrer, 2003), leucine-rich repeats (Stumpp, 2003), and armadillo repeats (Parmeggiani, 2008), and used to generate combinatorial libraries containing

(22)

HER2 (Zahnd, 2007) have been selected by ribosome display and are denoted designed ankyrin repeat proteins, DARPins.

Affibody molecules are small three-helix bundle proteins with a binding surface made up of residues spread in helices I and II. This class of affinity proteins has been used in all the studies in this thesis and I have therefore devoted a whole chapter (2.3.4.) to describe them in detail.

Table 1. Examples of affinity proteins derived from non-immunoglobulin scaffolds

Class of affinity protein

Structure of binding surface

Size (aa) Number of disulfides

Chemical synthesis

Kunitz domains Single loop 58 3 Not reported

Thioredoxin-derived peptide aptamers

Single loop 108 1 Not reported

Cystine-knot miniproteins

Single loop 28-34 3 Le Nguyen, 1989

Daly, 1999

Anticalins Loop rich 160-180 0-2 Not reported

AdNectins Loop rich 94 - Not reported

DARPins Secondary structure 166 - Not reported

Affibody molecules Secondary structure 58 - Nord, 2001 Engfeldt, 2005

2.3.4 Affibody molecules

The Affibody scaffold has been derived from staphylococcal protein A, (SPA), a receptin found in the cell wall of Staphylococcus aureus where it interacts with immunoglobulins (Langone, 1982). SPA consists of five homologous domains, E, D, A, B and C, all of which bind to the Fc region of human IgG (IgG1, 2 and 4, but not IgG3) and also to the Fab part of antibodies from the human VH3 subclass (Jansson, 1998). SPA also binds to immunoglobulin molecules from several other species (Lindmark, 1983). These interactions have been extensively used for affinity purification of antibodies and in different types of immunoassays. The B domain of SPA has been used to derive the Z domain, which is the scaffold that Affibody molecules originate from (Figure 3). The Z domain differs from the B domain in that it has a Gly to Ala substitution in position 29 for increased stability to hydroxylamine. Moreover, Ala¹ has been

(23)

substituted for Val for subcloning reasons (Nilsson, 1987). Upon mutating residue 29, the Fab binding capacity was significantly reduced (Jansson, 1998). Z is made up of a single polypeptide chain of 58 amino acids with no cysteines, which folds rapidly (Arora, 2004) and independently into an anti-parallel three-helix bundle structure. The Fc-binding has been located to 11 surface-exposed residues distributed over helices I and II (Deisenhofer, 1981).

The strategy from which Affibody molecules have spurred relies on the idea to introduce amino acid substitutions at specific sites in order to displace the native binding specificity but preserve the structural folding. 13 residues, including the 9 of the 11 Fc-binding residues, along with 4 additional residues located at the same face, were selected for genetic randomization to create a library of Affibody variants (Nord, 1995) (Figure 3). The first Affibody molecules with novel binding specificities were generated by incorporating a library of ~10⁷ members into a phagemid vector allowing for phage display and selection against Taq DNA polymerase, human apolipoprotein A-1 variant, and human insulin (Nord, 1997). It was shown that the IgG interaction had been replaced with the novel binding specificities. Affibody molecules have since been selected to bind target molecules of various size, shape and origin in different applications, reviewed by Nygren (Nygren, 2008).

The binding affinity of Affibody molecules selected from the initial library was typically in the low micromolar range. Strategies to improve the affinity of those binders by affinity maturation have been successful in several cases, reaching equilibrium dissociation constants, KD, in the nanomolar range (Gunneriusson, 1999) (Nord, 2001). However, by increasing the library size to ~10⁹ members, high affinity binders (nanomolar KD) could be selected directly from naive libraries. Examples of such Affibody molecules are the ones targeted to HER2 (Wikman, 2004), transferrin (Gronwall, 2007b), amyloid beta peptide (Gronwall, 2007a) and EGFR (Friedman, 2007). Directed evolution strategies have subsequently generated picomolar binders interacting with e.g. the HER-2 receptor (KD=22 pm) (Orlova, 2006), which are used in papers III-VII of this thesis. Also the EGFR-binder has been subjected to affinity maturation, resulting in a 30-fold increase in affinity (Friedman, 2008).

(24)

Figure 3. Schematic overview of the IgG binding domains of staphylococcal protein A, the Z domain, and the Affibody scaffold

The structure of Affibody molecules has been studied using NMR and X-ray crystallography.

These studies have been accompanied by circular dichroism experiments and reveal that most selected Affibody variants fold into a compact three-helix bundle structure similar to the parental Z domain, despite the rigorous amino acid substitutions induced by the randomization, that is affecting more than 20 % of the residues. High-precision solution structures, solved by NMR, of the Affibody molecules ZTaq and anti-ZTaq, and their complex (ZTaq:anti-ZTaq), reveal well-folded three-helix bundles, both in their free forms and in co- complex (Lendel, 2006). However, the Affibody molecule ZSPA-1, with lower target binding affinity, has a molten globule character in solution and only folds into the three-helix bundle structure upon binding to its target, Zwt (Lendel, 2002). A solution structure of an Affibody- dimer in complex with its target amyloid Ƣ-peptide, AƢ(1-40), have revealed an unexpected structural topology, in which the N-terminus of the Affibody molecules forms a beta strand, that can form hydrogen bonds with the AƢ-peptide (Hoyer, 2008).

(25)

Affibody molecules have proven to have potential for use in several applications, ranging from bioseparation, diagnostics and viral targeting, to therapy. One important feature of the Affibody scaffold is that it is amenable to chemical synthesis protocols, described in chapter 3.3. Its small size (58 aa) allows for total synthesis with respectable yields, and the high solubility and independent folding mechanism allows for facile recovery of correctly folded proteins. Peptide synthesis brings measures for chemical modifications, such as incorporation of non-coded residues, site-specific introduction of functional probes or other features, that might be difficult to achieve by recombinant production routes.

The small size of Affibody molecules is an advantageous feature that has been employed also for in vivo applications based on molecular recognition. The rather large size and the presence of an Fc domain can be limiting factors for full-sized antibodies in applications such as in vivo molecular imaging due to long circulation times and poor tissue penetration (see chapter 4.3.7 for further details). Affibody molecules directed to tumor-associated antigens have been studied as alternative targeting agents (Nordberg, 2008, Steffen, 2006). Particularly, a HER2- binding Affibody has been extensively characterized in several preclinical and pilot clinical studies (Baum, 2006, Feldwisch, 2006, Orlova, 2007a). Paper III-VII of this thesis describes how Affibody molecules have been engineered to serve as targeting agents in in vivo imaging of HER2-expressing tumors.

(26)

3. Protein and peptide chemistry

3.1 Peptide and protein production

There are three principally different ways in which peptides and proteins can be obtained;

either by extraction from natural sources, through recombinant production technologies, or by peptide synthesis.

Extraction of peptides and proteins from natural sources, such as plants, microorganisms, marine species or mammalian tissues has traditionally been a valuable method to obtain proteins. One important example of a therapeutic protein that was originally obtained from a native source is insulin, which is used for treatment of diabetes mellitus. Protein extraction from natural sources suffers from several drawbacks. The availability is limited, especially human organs are scarce, and the necessary protein purification methods are costly procedures.

Furthermore, the risk of transmission of viral and prion diseases through the use of extracts from animals or humans limit the application of proteins derived from such sources.

The development of the recombinant DNA technology has provided a means for cheap, large scale recombinant production of proteins. Recombinant protein production has been of great value both to increase the understanding of the nature of proteins and in applied research.

Two Nobel prize-rewarded achievements have been crucial for the development of recombinant DNA technologies, the discovery of restriction enzymes (Linn, 1968) and the invention of the polymerase chain reaction, PCR (Saiki, 1985). PCR is a powerful method for multiplication of DNA segments, whereas restriction enzymes and ligase allows for precise cutting and ligation of DNA fragments. The combination of the two technologies facilitates the introduction of protein-coding genes to host organisms for recombinant protein production.

In recombinant protein production strategies, the protein synthesis machinery of cells is employed to produce proteins from engineered DNA sequences. There are numerous recombinant DNA technologies that can be used for isolation and incorporation of a protein- coding gene into a host cell. The cells can originate from a wide range of organisms such as

(27)

bacteria, yeast, insects or mammals. A huge diversity of strategies for recombinant protein production has been reported, as well as down-stream processing protocols to obtain the pure protein (Gräslund, 2008). Although most proteins have the potential to be produced by recombinant technologies, certain types of proteins are not well suited for such routes. For instance, it is difficult to produce proteins that are toxic to the cell. Furthermore, membrane proteins and other poorly soluble proteins might be difficult to express successfully.

Another technology that has earned its inventor the Nobel Prize is solid phase peptide syntheis, SPPS (Merrifield, 1963). Peptide synthesis is an alternative to recombinant protein production routes, available to peptides and small proteins. In peptide synthesis, methods developed in the field of organic chemistry are applied to assemble polypeptides in an artificial manner. This strategy is discussed in further detail in chapter 3.3.

Depending on the production pathway, different options arise for modification of the peptides and proteins. The possibility to incorporate functional probes has been valuable to expand the applications of natural, recombinant and synthetic polypeptides.

3.2 Modifications of peptides and proteins

Peptides and proteins can be equipped with functional groups to make them better suited for a certain application. This has been an important strategy to expand the means in which polypeptides can be used, both as research tools but also in the clinic. Chemical modifications of different character can be realized through a diverse set of strategies commonly referred to as bioconjugation. In essence, bioconjugation is the process of coupling two biomolecules together via a covalent link.

Natural and recombinant proteins present several reactive groups that can be subjected to

(28)

modified by various reagents including active esters. Most often there are many amino groups in a protein, and it might be difficult to discriminate between them using amine-reactive reagents, which results in a heterogeneous mixture of modified species. Some of the reporter groups described below requires very precise positioning. For instance, certain probes respond to minor changes in the microenvironment and therefore require atomic accuracy in the conjugation method in order to realize its full potential. Such correctness might be difficult to accomplish by bioconjugation to amino groups, but can be addressed through chemical synthesis.

Figure 4. Schematic view of a protein surface with some representative functional groups

There is a plethora of different probes that can be used to modify the structure and function of peptides and proteins. One important class of reporter groups are those that can be used in protein detection assays, both for biotechnological and diagnostic applications. Affinity probes are widely used in biotechnology, as e.g. ligands in affinity chromatography to purify proteins, but also as immobilization handles to attach proteins onto solid supports. Biotin is a small hapten with a high-affinity interaction with streptavidin that has been extensively used as an affinity probe for many purposes. One important application has been to immobilize proteins onto surfaces, such as microarray slides. The labeling of proteins and peptides with fluorescent probes has been extensively used for various protein detection assays. The ability to fluoresce can be given a protein either through coupling of a fluorescent probe, or through fusion with a fluorescent protein such as the Nobel Prize-awarded green fluorescent protein, GFP (Chalfie, 1994) (Tsien, 1998). In particular, fluorescent probes have played a central role in the

(29)

development of protein microarrays. Fluorescence is further discussed in chapter 4.2.1.

Another type of reporter groups that facilitate detection is molecular structures that are capable of coordinating radionuclides. Such chelators can be introduced to generate radiopeptides for various protein binding assays, including ELISA, but has also been of significant importance in the development of tracer molecules for diagnostic in vivo detection assays such as molecular imaging (molecular imaging is further described in chapter 4.3.). An example of probes that can be used to facilitate detection in biophysical studies of proteins is spin labels, that provides a means for studying peptide conformations in solution by electron spin resonance (ESR) (Marchetto, 1993). Another interesting type of chemical modification is polymer conjugation. Addition of polyethyleneglycol, PEG, is commonly used as a strategy to prolong the half-life of therapeutic peptides and proteins in vivo. PEGylation has been found to not only improve the circulation time, but bring additional benefits in terms of physical and thermal stability, resistance to proteolytic degradation, solubility and immunogenicity (Bailon, 2001).

Bioconjugation offers several valuable methods to modify proteins, however its scope of use is limited to the decoration of an already existing peptide backbone and cannot be applied for sequential integration of non-coded amino acids. A general strategy for introduction of non- natural amino acid derivatives at specific sites during recombinant production of proteins is offered by the nonsense suppression mutagenesis method (Noren, 1989). The method involves site-directed mutagenesis to introduce an amber stop codon at a site of interest, and in vitro aminoacylation of the corresponding tRNA. The non-natural amino acid is thereafter incorporated into the polypeptide chain by means of the ribosomal translation machinery. In a way, nonsense suppression mutagenesis can be regarded as a means to expand the genetic code beyond the use of the 20 natural amino acids. An important development in suppressor tRNA techniques has been the incorporation of amino acids with side chains carrying unique chemical groups for bioconjugation. Two examples are keto-carboxyl-containing amino acids for hydrazone or oxime bond forming modifications, and amino acids with azido functional groups for generation of triazole derivatives (Dieters, 2003). Many amino acid derivatives have

(30)

specific antibodies. Gene fusion tags that allow for specific modifications inside cells have been reported, such as the Halotag, and the biotag. The Halo tag is based on a modified bacterial haloalkane dehalogenase and can be used for covalent bond formation with chloroalkane-fuctionalized probes (Los, 2008). The biotag consist of a biotin acceptor peptide fused to the gene for Escherichia coli biotin ligase (BirA) and can be used for producing biotinylated recombinant proteins in mammalian cells (Kulman, 2007).

Bioconjugation, genetic engineering and nonsense suppression offer valuable methods to modify functions of peptides and proteins, although they do not provide the operational freedom required for realization of modification patterns of complex or sensitive nature. As an alternative, organic synthesis is not limited to the use of natural or DNA-coded building blocks and offers means to introduce reporter groups and reactive probes in a site-specific manner, with the only assumption that they have to withstand the employed reaction conditions.

3.3 Chemical synthesis

3.3.1 History and development of Solid Phase Peptide Synthesis

The field of peptide chemistry begins with the synthesis of short amino acid oligomers by Emil Fischer in 1901 (Fischer, 1901) and Theodor Curtius in 1902 (Curtius, 1902). It was also then that the word peptide was first introduced, as Fischer named his product glycyl-glycine a dipeptide. A few years later, Fisher reported the synthesis of an octadecapeptide using acid chlorides for peptide bond formation (Fischer, 1907). Some 30 years later, Bergman and Zervas introduced the benzyloxycarbonyl (Cbz or Z) group as an amine protecting group that could be removed without cleavage of the peptide bonds, thus enabling controlled N-terminal chain elongation (Bergmann, 1932). The first biologically active synthesized peptide was oxytocin, reported by Du Vigneaud in 1953 (Du Vigneaud, 1953), for which he was later awarded the Nobel Prize in Chemistry. Although improved techniques and new reagents allowed for successful synthesis of an impressive number of small peptides, technical difficulties, such as poor solubility of protected peptides and slow reaction rates hampered the realization of synthesis of longer polypeptides. Also, purification of the product from by- products became a major issue as the number of residues increased.

(31)

A new era for peptide synthesis began in 1963, when Bruce Merrifield introduced the concept of solid phase peptide synthesis, (SPPS). The Nobel Prize-awarded technology relies on the anchoring of the C-terminal amino acid to a solid polymer support, and the subsequent coupling and deprotection of N-protected amino acids in a stepwise manner and release from solid support upon completed synthesis (Figure 5). The first report of this technology was the synthesis of a tetrapeptide using chloromethylated polystyrene as solid support, carbobenzoxy- protected amino acids as building blocks and deprotection with strong acid (Merrifield, 1963).

The major achievement was that reagents and impurities could be removed by washing and filtration, thus allowing for excess use of reagents to drive the coupling reactions to completion. When Merrifield later introduced the t-butyloxycarbonyl (Boc) group as an amine protecting group and used a benzyl ester as a linker to the solid phase, milder reaction conditions could be used (Merrifield, 1964), allowing for synthesis of longer sequences, such as insulin (Marglin, 1966), and the 124 aa bovine pancreatic ribonuclease A (Gutte, 1971).

The repetitive acid treatments during syntheis with the Boc/benzyl strategy, and the use of strong acid in the final deprotection step caused a number of undesired side reactions. It became evident, that for biologically active proteins to become readily attainable by synthetic means, even milder reaction conditions were required. The base-labile 9- fluorenylmethyloxycarbonyl (Fmoc) protection group was introduced by Carpino and Han in 1972 (Carpino, 1972) and realized for solid phase synthesis by Atherton and coworkers (Atherton, 1978a, Atherton, 1978b). By using Fmoc as a temporary amino protecting group in combination with acid labile tert-butyl derivatives for side chain protection, a complete differentiation in reaction conditions for the different deprotections was achieved, which was not possible when using acidic reagents of varying strength. This chemical differentiation is often referred to as orthogonality. In the context of synthetic chemistry, an orthogonal system is defined as a set of completely independent classes of protecting groups, where each class of groups can be removed in any order and in the presence of all other classes (Barany, 1977).

Orthogonality is further discussed in chapter 3.3.3.

(32)

Figure 5. The principles of solid phase peptide synthesis

(33)

3.3.2 Principles of Fmoc SPPS

The underlying principle of most peptide bond-forming coupling methods is the activation of the carboxyl group with an electron-withdrawing leaving group that renders the carbon atom susceptible for nucleophilic attack by the amino group. The activating group or reactions should be carefully chosen in order to achieve high reaction efficiency and at the same time avoid side reactions. There are several different coupling methods employed in solid phase synthesis. Traditionally, carbodiimides have been used as in situ activating reagents (Sheehan JACS 1955). Carbodiimides have also been used to generate pre-activated symmetrical anhydrides. The addition of triazolols, such as 1-hydroxy-benzotriazole (HOBt) or 1-hydroxy- 7-aza-benzotriazole (HOAt) in the pre-activation of Fmoc-amino acids with carbodiimides was introduced later and found to result in active esters with slightly reduced reactivity and reduced propensity for racemization. More recently, in situ activating reagents have emerged that totally omit the use of carbodiimides, such as the uronium or phosphonium-based salts HBTU, HATU and PyBOP. In situ activating coupling reagents are the most widely used procedure today, being suitable for use in automated synthesis and giving fast reactions also for sterically hindered amino acids.

Using any of the activation methods above, amino acids are coupled sequentially in a stepwise manner to amino groups presented on the resin. In order to allow for unambiguous peptide bond-formation, all reactive sites but the carboxyl group of the amino acid has to be masked.

Not all amino acids require side chain protection groups, but certain reactive groups need special attention. The alcohol functions of serine and threonine, the phenol of tyrosine, the guanidine of arginine and the thiol of cysteine can all undergo acylation and alkylation and need protection during the synthesis to prevent the formation of branched peptides. Histidine can undergo both acylation and racemization if not protected properly. The carboxyl groups in the side chains of aspartic and glutamic acids must be protected in order to achieve unambiguous activation. Asparagine and glutamine might be used without protection of the primary amide functionalities, however dehydration during activation can be avoided by use of